Ocular Cellscope

Christopher EchaniqueBjörn Hartmann, Ed.Daniel Fletcher, Ed.

Electrical Engineering and Computer SciencesUniversity of California at Berkeley

Technical Report No. UCB/EECS-2014-91

http://www.eecs.berkeley.edu/Pubs/TechRpts/2014/EECS-2014-91.html

May 16, 2014

Copyright © 2014, by the author(s).All rights reserved.

Permission to make digital or hard copies of all or part of this work forpersonal or classroom use is granted without fee provided that copies arenot made or distributed for profit or commercial advantage and that copiesbear this notice and the full citation on the first page. To copy otherwise, torepublish, to post on servers or to redistribute to lists, requires prior specificpermission.

Acknowledgement

I am thankful for the help of my fellow masters students working on thisproject: PJ Loury, Doug Webster, and Naya Loumou. I am also thankful forthe generous help of my advisors, including Frankie Myers, Clay Reber,Dan Fletcher, Robi Maamari, Tyson Kim, Michael Yen, Todd Margolis, andBjoern Hartmann.

 

University of California, Berkeley College of Engineering

MASTER OF ENGINEERING - SPRING 2014

EECS

Visual Computing and Computer Graphics

Ocular Cellscope

Christopher Echanique

This Masters Project Paper fulfills the Master of Engineering degree requirement.

Approved by:

1. Capstone Project Advisor:

Signature: __________________________ Date ____________

Print Name/Department: Bjoern Hartmann/EECS

   2. Faculty Committee Member #2:

Signature: __________________________ Date ____________

Print Name/Department: Daniel Fletcher/Bioengineering

 

 

     

Ocular  CellScope    

Masters  of  Engineering  Final  Report                          

Christopher  Echanique  

 

Acknowledgements    

I  am  thankful  for  the  help  of  my  fellow  masters  students  working  on  this  project:  PJ  Loury,  Doug  Webster,  and  Naya  Loumou.  

 I  am  also  thankful  for  the  generous  help  of  my  advisors,  including  Frankie  Myers,  Clay  Reber,  Dan  Fletcher,  Robi  

Maamari,  Tyson  Kim,  Michael  Yen,  Todd  Margolis,  and  Bjoern  Hartmann.  

                                 

This  report  covers  some  of  the  work  done  by  the  master’s  students  on  the  Ocular  CellScope  Project.  The  methods  and  

results  of  mobile  application  design  will  be  emphasized  because  it  was  the  work  directly  performed  by  the  author.  

 

 

Abstract  

Globally,  314  million  people  are  visually  impaired,  with  87%  of  this  affected  

population   residing   in   developing   countries   [1].   Glaucoma,  macular   degeneration,  

and   diabetic   retinopathy   are   all   leading   causes   of   blindness   in   the   United   States.  

These   diseases   require   routine   eye   exams   for   their   effective   diagnosis   and  

management.  Identifying  eye  diseases  at  an  early  stage  is  crucial  to  preventing  the  

condition   from  worsening   and   causing   permanent   vision   loss.   However,   the   high  

cost  of   retinal   imaging  equipment   is  often  prohibitive   for  many  clinics  around   the  

world  and  many  people  forgo  necessary  eye  exams.  Thus,  low-­‐cost  handheld  retinal  

imaging  tools  which  could  be  used  at  home,  in  a  primary  care  clinic,  or  a  pharmacy,  

would   dramatically   expand   access   to   eye   examinations.   We   present   Ocular  

CellScope,   a  mobile  ophthalmoscope  device   that   leverages   smartphone   technology  

to  effectively  capture  retinal  images  and  manage  patient  records.  Our  system  is  the  

second  design  iteration  of  the  existing  prototype,  addressing  a  number  of  usability  

issues   associated   with   the   lack   of   an   integrated   mobile   application   and  

ergonomically   friendly  design.   It   features   an   advanced  optical   lens   attachment   for  

the  iPhone  built  with  a  custom  illumination  system.  A  mobile  application  has  been  

designed   to   capture,   store,   and   review   retinal   images,   as   well   as   record   patient  

information.   Our   device   is   capable   of   producing   high-­‐quality   retinal   images  

comparable  to  that  of  high-­‐end  retinal  imaging  equipment  currently  on  the  market  

and  is  priced  at  a  significant  fraction  of  the  price  of  current  competitors.  

 

Introduction  

The  diagnosis  and  monitoring  of  retinal  diseases  such  as  diabetic  retinopathy,  

macular   degeneration,   and   CMV   retinitis   has   been   a   key   focus   in   the   field   of  

ophthalmology.  The  use  of  medical  equipment  with  retinal  imaging  functionality  has  

been   a   significant   factor   in   the   diagnosis   of   these   eye   diseases.   Since   the  

introduction  of   such  equipment   in   the   late  19th   century,   technical   improvement  of  

retinal   imaging   has   been   steady   [1].   Retinal   imaging   devices   known   as   fundus  

cameras   are   most   commonly   used   in   ophthalmology   clinics   to   provide   a   full  

evaluation  of  a  patient’s  retinal  health.  However,  despite  the  extensive  functionality  

of   these  devices,   their   limitations   including   cost,   size,   and   accessibility  have   given  

rise   to   the   emergence   of   different   imaging   methods   that   address   some   of   these  

issues.   These   new   advanced   ophthalmoscopes   aim   to   provide   retinal   imaging   on  

mobile  platforms  at  low  costs,  yet  trade  off  the  advanced  functionality  of  traditional  

fundus   cameras.   The   goal   of   Ocular   CellScope   is   to   bridge   this   gap   between   the  

functionality   of   fundus   cameras   and   the   low   cost   of  mobile   ophthalmoscopes   that  

seems  to  be  lacking  in  the  competitive  landscape.  Our  mobile  phone-­‐based  device  is  

the  second  design   iteration  of  Ocular  CellScope,  which   leverages   the  compact  size,  

high-­‐resolution   camera,   large   data   storage   capacity,   and   wireless   data   transfer  

capabilities   of   current  mobile   devices   to   capture   retinal   images   for   diagnosis   and  

monitoring.  The   features  of  Ocular  CellScope  challenge  competing  products  on   the  

market  and  our  deployment  of  this  device   into  the  market  of  developing  countries  

such   as   Thailand   will   help   target   a   need   for   these   low-­‐cost   fully   featured   retinal  

imaging  devices.  

 

Literature  Review  

The  current  gold  standard  in  retinal   imaging  is  the  tabletop  fundus  camera.  

These   stationary   devices   found   in   ophthalmology   clinics   offer   a   wide   array   of  

advanced   features   for   imaging   the   fundus   of   the   eye.   Such   features   include   high-­‐

resolution   retinal   images,   non-­‐mydriatic   imaging,   large   retinal   viewing   space   (45  

degrees),   software   interface   for   file   storage   and   image   analysis,   and   advanced  

software  functions  like  auto-­‐focus,  auto-­‐stabilization,  and  image  stitching.  Dozens  of  

fundus  camera  models  have  been  introduced  to  the  market.  Among  the  most  widely  

used  devices  are  the  Canon  CR-­‐2,  the  Centervue  DRS,  and  the  Nidek  AFC,  which  offer  

their  own  set  of  features  in  addition  to  the  ones  above.  The  Canon  CR-­‐2  model  offers  

digital  filter  processing  to  enhance  screening  exams,  and  low  flash  photography  for  

improved   patient   comfort,   and   a   lighter   weight   (33   lbs.)   design   for   increased  

portability   [2].   The   Centervue   DRS   model   also   offers   a   fully   automated   design  

requiring  minimal  operator  skill,  web  connectivity  for  image  transfer  to  a  network,  

and  software  features  for  auto-­‐sensing,  alignment,  and  focus  [3].  The  Nidek  AFC  is  

one   of   the   most   comprehensive   devices   in   that   it   offers   all   of   these   features   in  

addition   to  an   “autostereo”   feature,  which  provides  a   serial  analysis  useful   for   the  

diagnosis  of  diseases  such  as  Glaucoma  [4].  Despite   the   large  number  of  advanced  

features   available,   fundus   cameras   suffer   from   several   drawbacks.   Cost   is   the  

primary   limitation,  with  most   of   these   devices   ranging   from   $10,000-­‐$20,000   per  

unit.   This   price   point   is   significantly   higher   than   other   types   of   retinal   imaging  

devices   on   the  market.  Also   these  devices   are  much   larger   and  heavier   than   their  

mobile   counterparts,   requiring   a   stationary   setup   for   patients   to   sit   upright   for  

 

exams.   This   can   pose   issues   for   sick   and   hospitalized   patients.   These   drawbacks  

make   it   difficult   to   compete   in   a   market   focused   on   low   cost   mobile   retinal  

screenings  as  well  as  limit  its  presence  in  markets  of  developing  countries.  

Mobile   ophthalmoscopes,   on   the   other   hand,   have   entered   the   market   in  

recent   years   to   provide   a   low   cost   portable   alternative   to   expensive   stationary  

fundus   cameras.   These   devices   can   be   categorized   into   two   types:   the   standard  

ophthalmoscope   and   the   mobile   digital   imaging   ophthalmoscope.   The   standard  

ophthalmoscope   is  a  handheld  device  with  an  optic   lens  and   illumination  used   for  

simple   viewing   of   the   retina.   One   such   device   is   the   Keeler   Standard  

Ophthalmoscope,  which  features  an  adjustable   focusing   lens,  an  aperture  selection  

disc   for  dilated  and  undilated  pupils,  and  a  red-­‐free   filter   for  easy   identification  of  

retinal  features  [5].  These  devices  have  the  lowest  costs  on  the  market,  with  prices  

ranging  from  $100-­‐$200  due  to  their  simple  design  and  basic  functionality.  However  

these  devices  are  not  as  practical  for  retinal  screening  since  they  lack  digital  retinal  

image   storing   and   thus   require   a   trained   physician   to   perform   the   exam   and  

diagnosis.   The   second   type   of   device   mentioned   is   the   mobile   digital   imaging  

ophthalmoscope,  which  categorizes  our  Ocular  CellScope  product.  The  newest  of  all  

types   of   retinal   imaging   device   types   available,   this   type   has   the   least   number   of  

models  on  the  market.  The  largest  competitor  to  Ocular  CellScope  is  the  Welch  Allyn  

iExaminer  System.  This  system  features  a  mobile  PanOptic  Ophthalmoscope  with  an  

iPhone   adapter   and   integrated   iExaminer   App   to   capture,   send,   and   store   retinal  

images  as  well  as  record  patient  information  [6].  The  device  has  a  25-­‐degree  field  of  

view  and  a  patented  anti-­‐glare   system   to   reduce  unwanted  glare   in   images  and   is  

 

valued  at  $1000.  The  ease  of  use,  portability,  and  cost  of  this  device  make  it  practical  

for   use   in   retinal   disease   screenings,   however   this   systems   faces   a   number   of  

drawbacks.   The   main   limitation   is   the   25-­‐degree   field   of   view   it   provides.  

Traditional   fundus   cameras   offer   45   degrees,   which   is   at   times   necessary   for   a  

complete  diagnosis  as  some  diseases  depend  on  the  outer  regions  of  the  retina.  Also  

this   system   lacks   a   low-­‐intensity   illumination   option   for   patient   comfort   during  

exams   and   offers   no   hardware   controls   or   web   connectivity   in   the   iPhone  

application.  

The   Ocular   CellScope   device   aims   to   address   these   limitations   associated  

with  both  stationary  fundus  cameras  and  mobile  ophthalmoscopes  available  on  the  

market.  The  device  provides  an  attachment  to  the  iPhone  similar  to  the  Welch  Allyn  

iExaminer  system.  However,  the  device’s  illumination  system  will  be  fully  integrated  

with  the  iPhone  via  Bluetooth  to  allow  user  control.  The  system  also  features  a  set  of  

fixation  lights  to  capture  designed  regions  of  the  retina  to  provide  a  retinal  field-­‐of-­‐

view  of  approximately  55  degrees   [1].  This   is  a  drastic   improvement   from  the  25-­‐

degree   field   of   view   offered   by   the   iExaminer   system.   In   addition,   the   mobile  

application  allows  the  user  to  not  only  store  patient  information  and  retinal  images  

but  also  upload  this  data  to  a  web  server  for  access  by  designated  physicians.  These  

added  features  will  compete  with  the  functionality  of  stationary  fundus  camera  and  

existing  mobile   ophthalmoscopes,   but   at   a  price  point   on   the  order  of   $1000.  The  

device   will   allow   for   a   completely   portable   and   inexpensive   solution   for   retinal  

imaging   that   could   be   used  both   in   the   hospital   setting   and   for   community   vision  

screening.  

 

Methods  

1.  Initial  Prototype  

The  initial  Ocular  CellScope  prototype  (shown  in  Figure  1)  consists  of  a  3D-­‐

printed  enclosure  housing  the  optic  lens  and  illumination  system,  which  is  powered  

by  a  9V  battery.  A  slot  designed  to  fit  the  iPhone  4/4s  models  aligns  the  back  camera  

of   the   iPhone  with   the   optic   lens   of   the   device   used   to   enhance   the   retinal   image  

view.  A  switch  on  the  back  side  of  the  device  toggles  power  to  the  white  LED  of  the  

illumination  system  and  a  knob  adjacent  to  the  switch  controls  the  intensity  of  the  

light.  Retinal  images  are  captured  using  the  native  iPhone  camera  application  while  

the  light  switch  is  powered  on  to  illuminate  the  retina.  

 

Figure  1.  First  Ocular  CellScope  Prototype  

2.  Prototype  Feedback  

Field-­‐testing   the   initial   Ocular   CellScope   prototype   in   Thailand   unveiled   a  

number   of   usability   issues.   These   issues   mainly   stemmed   from   the   lack   of   an  

integrated  mobile  application,  which  made  common  ophthalmologic  screening  tasks  

difficult  for  the  clinical  technicians  using  the  device.  The  lack  of  a  mobile  application  

 

forced   technicians   to   rely   on   the   iPhone’s   native   camera   application   to   capture  

retinal  images  with  the  initial  prototype.  This  approach  was  troublesome  due  to  the  

fact   that  all   images   from   the  native   camera  application  are   stored   in  a   centralized  

photo  library  on  the  phone,  making  it  difficult  to  parse  by  patient.  Technicians  took  

images   of   the  patient’s   paper  medical   record   to  mark   the  beginning  of   the   retinal  

image  set  for  that  exam  screening.  They  expressed  frustration  in  trying  to  separate  

patient  data  for  review.  Additionally,  the  bulky  nature  of  the  prototype  device  made  

it   difficult   for   technicians   to   operate   with   one   hand.   Since   the   native   camera  

application   required   the   user’s   touch   to   capture   images,   this   led   to   awkward  

repositioning  of  the  device  during  image  capture  and  often  times  resulted  in  blurry  

images   as   clinicians   readjusted.   These   poor   quality   images   were   also   due   to   the  

inability  to  lock  the  iPhone  camera’s  focus,  causing  the  autofocus  feature  to  result  in  

nondeterministic   image   acquisition   times.   Furthermore,   the   technicians   lacked   a  

platform  to  digitally  record  patient  information  and  link  this  to  the  set  of  acquired  

images.    

Patients  being  screened  also  issued  a  number  of  complaints  with  the  device.  

Many   patients   expressed   discomfort  with   the   intensity   and   duration   of   the  white  

LED   upon   image   acquisition.   The   hard   switch   design   of   the   illumination   system  

made  it  difficult   for  technicians  to  rapidly  turn  the   light  on  and  off   to  simulate  the  

speed  of  a  camera  flash,  exposing  patients  to  the  intense  white  light  for  longer  than  

tolerable   periods.   In   addition,   the   asymmetric   L-­‐shaped   design   of   the   device  

obstructed   the   patient’s   face   during   retinal   imaging   of   the   left   eye   as   part   of   the  

device  enclosure  extended  toward  the  patient’s  nose.  

 

3.  Second  Design  Iteration  

Hardware  

The   feedback   obtained   from   technicians   and   patients   has   helped   guide   the  

redesign   of   the   device   hardware   for   the   second   prototype   iteration   of   the   Ocular  

CellScope   device.   The   hardware   redesign   focuses   on   three   main   subsystems:   the  

illumination   system,   the   fixation   light   system,   and   the   enclosure   redesign.   For   the  

illumination  system,  a  red  LED  is  incorporated  to  allow  technicians  to  focus  on  the  

retina   between   image   acquisitions,   allowing   for   improved   patient   comfort   due   to  

reduced  retinal  sensitivity.  In  addition,  an  external  fixation  light  system  is  added  to  

the  hardware  to  provide  patients  with  a  region  to  focus  their  eyes  during  screenings  

and   allow   for   a   wider   degree   of   retinal   imaging   space.   Lastly   the   design   of   the  

enclosure  is  enhanced  to  provide  a  more  ergonomic  feel  for  both  the  technician  and  

the  patient.  

Software  

User  feedback  from  the  initial  prototype  has  also  inspired  the  creation  of  an  

iPhone  application  adapted   for   the  device  hardware  and  user.  After  an  analysis  of  

the   competitive   landscape   for   ophthalmoscope-­‐based   mobile   applications,  

particularly   the  Welch   Allyn   PanOptic   iExaminer   application,   this   application   has  

been  designed  to  mimic  a  similar  workflow  for   image  acquisition  and  patient  data  

storage.   The   application   focuses   on   providing   a   structured   workflow   for   the  

technicians  to  follow  by  allowing  technicians  to  enter  patient  information,  capture  a  

set  of  retinal  images  for  each  eye,  and  persistently  store  this  data  on  the  phone  for  

later   access.   Features   have   been   added   to   mitigate   the   problem   of   poor   quality  

 

images,  such  as  locking  the  focus  during  image  acquisition  and  capturing  a  series  of  

images   over   short   intervals,   similar   to   that   of   the   iExaminer   application.   The  

primary   difference   between   the   two   is   that   the   Ocular   Cellscope   application   is  

integrated  with   the   illumination   and   fixation   light   system   through  Bluetooth.  This  

allows   the   phone   to   control   the   device’s   hardware   components   wirelessly   and  

alleviating  the  burden  of  manual  control,  a  feature  that  the  iExaminer  system  lacks.    

 

Results    

1.  Hardware  Solutions  

The   first  hardware  design  change   in   the  next  design   iteration  of   the  Ocular  

CellScope  was   the   addition   of   red   light   focusing   into   the   illumination   system.  The  

initial   Ocular   CellScope   prototype   uses   three  white   LEDs   to   illuminate   the   retina.  

This   type  of   illumination  outputs   light  at  a  very  high   intensity,   causing  substantial  

discomfort   in   patients.   Although   white   LED   is   necessary   to   capture   full   detail   in  

retinal  images,  it  is  important  to  reduce  the  duration  that  the  patient  is  exposed  to  

this   type   of   light   in   order   to   improve   comfort.   The   use   of   red   light   for   retinal  

focusing   in   between  white   light   flashes  during   image   acquisition   can   achieve   this.  

Red  light  is  much  more  comfortable  for  human  eyes  as  they  do  not  absorb  red  light  

as  easily  as  other  colors  in  the  visible  spectrum.  An  added  benefit  is  that  the  use  of  

red  light  minimizes  the  pupillary  response,  allowing  the  device  to  be  used  on  dark-­‐

adjusted  eyes,   instead  of   requiring   chemical  dilation.  Thus,   three  655nm  deep   red  

Luxeon   Rebel   LEDs   were   added   a   single   white   LED   in   the   original   illumination  

system  (see  Figure  2a).  This  wavelength  was  selected  because  it  is  long  enough  such  

 

that  the  human  eye  is  not  very  sensitive  to  it,  but  short  enough  such  that  it  will  still  

pass   through   iPhone  camera   infrared   filters.  Three  high-­‐powered  LEDs  were  used  

so  that  sufficient  red  light  would  pass  through  the  optical  system  to  illuminate  the  

eye  with  enough  contrast  for  the  camera  to  focus.    

   

Figure  2.  Illumination  System  with  Red  LEDs    

Second,   a   fixation   light   system   was   implemented   into   the   design.   Fixation  

lights   are   small   visible   targets   in   the   ophthalmoscope’s   viewing   space   where  

patients  can  fix  their  eyes  during  image  capture.  By  adding  five  lights  (four  in  each  

cardinal  direction  and  one  in  the  center),  the  device  can  provide  visual  cues  to  lock  

the  patient’s  focus  at  strategic  points  in  order  to  get  sufficient  coverage  of  the  retina  

for  diagnosis.  This  also  ensures  that  the  eye  is  not  moving  and  causing  motion  blur  

in   the   images.  Five  565nm  green  LEDs  were  selected   for   the   fixation   light   system.  

Green  was  chosen  because  human  eyes  are  highly  sensitive   to  green,  and   it  would  

contrast  with  red  focusing  illumination.    

Third,  the  form  factor  of  the  device  needed  to  be  streamlined.  The  old  model  

enclosure  is  awkward  to  handle,  at  times  requiring  the  technician  to  use  two  hands  

 

to   operate.   This   disrupts   the   workflow   of   retinal   imaging,   especially   when  

technicians  need  a  free  hand  to  operate  the  phone  or  open  the  patient’s  eye.  Also  the  

asymmetric  L-­‐shaped  design  of   the  original  enclosure   is   shaped  sometimes  blocks  

the   patient’s   eye   when   the   other   eye   is   being   imaged.   The   new   Ocular   CellScope  

enclosure   (shown   in   Figure   3)   has   a   simplified   ergonomic   design   inspired   by  

standard   ophthalmoscopes,   making   is   easier   to   hold   with   one   hand   and   more  

natural   for   technicians   to   operate.   In   addition,   the   external   light   switch   and   light  

intensity  knob  control  have  been  removed  in  this  new  design,  since  the  illumination  

will  now  be  control  through  the  mobile  application  interface.    

 

Figure  3.  Redesigned  CellScope  Enclosure    

2.  Software  Solutions  

The   mobile   application   has   been   designed   to   provide   technicians   with   a  

convenient  way  of   acquiring  patient  data   and  an   improved   interface   for   acquiring  

 

high  quality  retinal   images.  Since  the  device  enclosure  has  been  designed  to  house  

the  iPhone  4  and  4s  models,  software  development  has  been  targeted  to  these  two  

platforms,   though   the   mobile   application   is   forward   compatible   with   iPhone  

5/5s/5c  models  to  allow  for  future  upgrades.  

Patient  Data  Model  

Technicians  need  a  convenient  way  of  recording  and  storing  patient  data  at  

the  beginning  of  each  exam.  This  data  should  be  accessible  when  the  application  is  

closed  and  revisited.  To  achieve  this,  a  database  management  system  is  required  to  

ensure   persistent   storage   of   data   between   application   uses.   Apple   provides   two  

main   platforms   for   data   persistence:   SQLite   and  Core  Data.   SQLite   is   a   library   for  

relational   database  management   known   for   its   speed   and   reliability,   especially   in  

handling   sizable   sets   of   data.   Core  Data,   on   the  other  hand,   is   an  object   relational  

mapping  system  that  interfaces  with  SQLite,  yet  allows  for  the  direct  manipulation  

of  high  level  objects  as  opposed  to  raw  relational  data  [7].  The  main  drawback  here  

is   the   performance   costs   in   dealing   with   large   sets   of   data.   However,   since  

technicians  will  not  be  handling  considerably  large  sets  of  patient  data  locally  on  the  

phone,  the  Core  Data  framework  is  preferable  for  the  purposes  of  this  application.  

To  model  the  patient  exam  information  and  associated  images,  entities  were  

created   using   the   Core   Data   framework.   Each   entity   is   essentially   an   object   in  

Objective  C  containing  a  set  of  attributes  associated  with  it,  and  these  attributes  are  

assigned   a   specific   type   depending   on   the   kind   of   data   being   held.   Figure   4  

illustrates  the  entities,  attributes,  and  entity  relationships  used  for  this  application.  

An   “Exam”   entity   has   been   created   with   a   set   of   attributes   corresponding   to   the  

 

patient’s   name,   identification   number,   the   date   of   the   exam,   and   notes   to   record  

during   the   screening.   Similarly,   an   “EyeImage”   entity   has   been   created   with  

metadata  for  the  image,  including  the  type  of  eye  (left  or  right)  and  the  fixation  light  

triggered  to  capture  the  image.  The  “Exam”  entity  shares  a  one-­‐to-­‐many  relationship  

with   the   “EyeImage”   entity,   as   the   logic   follows   that   one   exam   contains   multiple  

images  associated  with  it.  

 

Figure  4.  Core  Data  Model  Graph  

It  is  important  to  note  that  the  raw  image  data  is  not  being  stored  in  the  data  

model   but   rather   a   file   path   to   this   image   data.   While   the   Core   Data   framework  

supports   the   storage   of   this   raw  data  within   attributes,   this   approach   is   not   good  

practice  due  to  the  size  of  these  high-­‐resolution  images.  To  ensure  efficient  memory  

usage  and  performance,   these   images  are  stored  on  the  phone  using  Apple’s  Asset  

Library  framework  and  accessed  with  the  file  path  saved  in  a  given  instance  of  the  

“EyeImage”   entity.   Lower   resolution   thumbnails,   however,   are   stored   in   the   data  

model   and   used   to   display   miniature   thumbnail   images   used   for   icons   in   the  

interface.  

 

Image  Acquisition  

One   of   the   biggest   concerns   expressed   by   the   technicians   using   the   initial  

prototype  was  the  inability  to  acquire  high  quality  retinal  images.  The  native  camera  

application  on  the   iPhone  placed  limits   image  acquisition  by  forcing  technicians  to  

snap   each   individual   retinal   image   as   they   held   the   device   with   both   hands.   The  

autofocus   feature   further   impeded   image   capture   as   the   camera   refocused   with  

slight   movements,   resulting   in   nondeterministic   image   capture   times   and   thus  

blurry   images.   Furthermore,   the   custom   lens   of   the   device   causes   images   to   be  

rotated  and  mirrored  when  viewed  on  the   iPhone  screen.  This  made  repositioning  

the   device   a   challenging   task   as  movements   had   to   be   interpreted   in   an   inverted  

manner.  Also  the  application  did  not  provide  a  way  to  revert  these  inverted  images  

into  their  original  state.  

The   goal   of   the   image   acquisition   feature   was   to   design   a   simple   tool   to  

capture  retinal  images  that  is  tailored  to  the  needs  of  the  technician.  Apple  provides  

two  frameworks  associated  with  image  capture  to  achieve  this.  The  UIImagePicker  

framework  manages  customizable  user  interfaces  for  taking  pictures  and  video  [8].  

This  method  is  essentially  a  black  box,  abstracting  much  of  the  camera  controls  from  

the  developer.  For  this  reason,  the  second  method,  the  AVFoundation  framework,  is  

preferred  as  it  provides  more  low-­‐level  controls  of  the  camera  device  such  as  focus  

and  exposure  lock  and  callback  functions  for  the  instant  an  image  is  captured.  

The  procedure  for  image  acquisition  has  been  designed  as  follows.  When  the  

user  selects  a  fixation  light  placeholder  on  the  images  tab  of  the  new  exam  view,  the  

application  segues  to  the  image  capture  view.  Here  the  phone  attempts  to  connect  to  

 

the   Ocular   CellScope   device   hardware   via   Bluetooth.  When   a   connection   is  made,  

phone  sends  a  signal   to   the  Bluetooth  device   to   trigger   the  corresponding   fixation  

light  and  the  low  intensity  infrared  light  for  retinal  illumination.  The  capture  button  

at   the   bottom   of   the   screen   initiates   the   image   acquisition.   In   order   to   provide  

robustness  of  the  system  in  the  presence  of  movement  by  the  user,  the  application  

captures   a   series   of   images   spaced   by   a   given   time   interval.   At   the   start   of   each  

image  in  this  series,  the  phone  sends  a  Bluetooth  command  to  briefly  turn  off  of  the  

infrared  and   fixation   light  and  trigger   the  white   flash   illumination.  After  all   retinal  

images   have   been   captured   for   the   given   fixation   light,   the   application   enters   an  

image  selection  view  where  the  user  can  then  review  and  select  the  images  to  save.  

Figure  5  below  demonstrates  this  workflow.  

 

  (a)   (b)   (c)   (d)   (e)    

Figure  5.  CellScope  Application  Workflow:  (a)  Main  screen,  (b)  patient  table  view,  

(c)  fixation  light  selection  view,  (d)  image  capture  view,  (e)  image  selection  view  

 

Conclusion  

  We   present   the   Ocular   CellScope,   a   mobile   ophthalmoscope   device   that  

leverages  smartphone  technology  to  effectively  capture  retinal  images  and  manage  

 

patient  records.  Our  system  is  the  second  design  iteration  of  the  existing  prototype,  

addressing  a  number  of  usability  issues  related  to  the  hardware  subsystems  and  the  

lack   of   an   integrated  mobile   application.   It   features   a   custom   illumination   system  

with   red   light   focusing   for   improved   patient   comfort   along   with   a   fixation   light  

system   to   provide   a   systematic   approach   for   full-­‐range   retinal   imaging.   A  mobile  

application   has   been   designed   to   capture   retinal   images   and   record   patient  

information,  as  well  as  control  with  device  hardware  components.  

The   fixation   light   system,   illumination   system,   enclosure   redesign,   and  

mobile  application  have  all  been  implemented  as  described  in  the  previous  sections.  

The  full  integration  of  the  system  has  yet  to  be  completed  and  tested  as  some  of  the  

ordered  hardware  components  required  for  integration  have  not  yet  been  delivered  

at  the  time  of  this  report.  Once  all  hardware  parts  are  acquired,  each  subsystem  will  

be  assembled  and  integrated  into  a  full  device  unit.  This  unit  will  be  evaluated  for  its  

effectiveness   in  acquiring  retinal   images,   the  quality  of   the  retinal   images,  and   the  

usability  of  the  device  itself.  

Future   work   on   the   Ocular   CellScope   device   involves   testing   the   device   in  

clinical  trials  to  determine  its  effectiveness  in  a  practical  setting.  Feedback  obtained  

from   technicians   and   patients   using   the   device   could   inspire   additional  

modifications   to   this   new   prototype.   Also   a   number   of   software   features   will   be  

considered  in  the  next  steps  of  the  device  design.  Adding  a  feature  for  technicians  to  

upload  patient  information  to  a  web  server  for  ophthalmologist  access  will  allow  for  

remote  diagnosis  of  potential  eye  diseases  in  patients.  To  improve  the  potential  for  

quick  onsite  diagnosis  of  eye  diseases,   the  use  of  computer  vision  to  automatically  

 

characterize   retinal   images   by   their   health   condition   will   also   be   considered   in  

future  iterations  of  the  device.    

 

 

 

References    [1] Maamari,  Robi,  Jeremy  Keenan,  Daniel  Fletcher,  and  Todd  Margolis.  "A  mobile  

phone-­‐based  retinal  camera  for  portable  wide  field  imaging."  British  Journal  of  

Ophthalmology  -­‐  (2012):  4-­‐12.  Copy  &  Paste  "Portable  Fundus  Camera."  

[2] "Canon  CR-­‐2  Digital  Non-­‐Mydriatic  Retinal  Cameras."  Canon  U.S.A.  :  Healthcare  

Technologies  :.http://www.usa.canon.com/cusa/healthcare/products/eyecare  

/digital_non_mydriatic_retinal_cameras/cr_2#Features.  

[3] "Fundus  Camera."  DRS  fundus  /  retinal  camera.  http://www.centervue.com/  

product.php?id=637.  

[4] "Autofocus  Fundus  Camera."  NIDEK  RSS.  http://usa.nidek.com/products/afc  

[5] "Standard  Ophthalmoscope."  Keeler.  http://www.keeler.co.uk/standard-­‐

ophthalmoscope-­‐4.htm.    

[6] "The  Welch  Allyn  iExaminer."  Welch  Allyn.  http://www.welchallyn.com/  

promotions/iExaminer/index.html.  

[7] Jacobs,  Bart.  "Data  Persistence  and  Sandboxing  on  iOS."  Code  Tuts+.  N.p.,  27  Dec.  

2012.  Web.  20  Mar.  2014.  <http://code.tutsplus.com/tutorials/data-­‐persistence-­‐

and-­‐sandboxing-­‐on-­‐ios-­‐-­‐mobile-­‐14078>.  

[8] “UIImagePickerController  Class  Reference."  UIImagePickerController  Class  

Reference.  Apple  Inc,  n.d.  Web.  20  Mar.  2014.  

<https://developer.apple.com/library/ios/documentation/uikit/reference/UII

magePickerController_Class/UIImagePickerController/UIImagePickerController

.html>.